Anthropic 申请历史性 IPO
Anthropic 正在申请史上规模最大的 IPO。这次 IPO 将触及每一位科技行业从业者。哪怕你从未购买过一股股票，这次 IPO 也至关重要。如果你为 Claude 付费，如果你在工作中使用它，或者如果你有用于投资的退休金，那么在几周内你很有可能会持有一部分 Anthropic 的股份。这不会是一次硬核的 IPO 财务分析，我们制作这个视频是为了揭示，对于每天使用并为 AI 付费的人群而言，这次 IPO 究竟会带来怎样的实际感受。互联网上到处都在尖叫着“800 亿美元的亏损”，但事实是他们并没有亏损任何东西。那么 Anthropic 为什么要进行 IPO？为什么是现在？Claude 会发生什么改变？如果你付费使用 Claude，该抱有什么预期？如果你决定投资，你投资的到底是一家怎样的公司？请看到最后，以了解关于这个真正具有历史意义的时刻所需知道的一切。让我们深入探讨。
英文原文
Anthropic is filing for the biggest IPO in history. And this IPO will touch everybody who works in tech. This IPO matters even if you never buy a single share. If you pay for Claude, if you use it at work, if you have retirement money that you invest, there's a good chance you'll own a piece of Enthropic within a few weeks. This isn't going to be a hardcore financial analysis of the IPO. We're making this video to shed light on what this IPO is going to feel like for people who use and pay for AI on a daily basis. The internet is screaming about $80 billion loss. But the truth is they didn't lose anything. So why is Anthropic going IPO? Why now? What will happen to Claude? What to expect if you pay for Claude? and what kind of company you're going to be investing in if you decide to invest. Stay till the end to understand everything you need to know about this truly historical moment. Let's dive in.IPO 解决的经济问题与庞大承诺
IPO 解决的经济问题是什么？当一家公司（任何公司）上市时，上市赋予了它私人投资根本无法提供的三样东西：全新的巨额资金池、按季度公布财务数据的义务，以及在股票市场交易的能力。上周，当 10 月 26 日发布 IPO 公告时，科技界爆发了各种反应和梗图，诸如：“他们到底怎么可能去 IPO？看看他们的债务。”现在，这些债务是真实存在的。《华尔街日报》报道了 6000 亿美元的承诺金额，S-1 申请文件本身显示了 7500 亿美元的承诺金额。让我们逐一解构它们。但在我们剖析之前，先简要回顾一下以便完全看清全局：Anthropic 运行在三大芯片家族之上——Google TPU、Amazon Trainium 和 Nvidia GPU。他们所有的承诺可以细分为三组：对 Google 的承诺、对 Amazon 的承诺以及对 Nvidia 的承诺。
英文原文
What economic problem does the IPO solve? When a company, any company, goes public, going public gives it three things that private investing simply can't. a new gigantic pool of money, a duty to publish their financials every quarter, and an ability to trade on the stock market. Last week, when the IPO announcement came out on October 26th, tech community blew up with reactions and memes along the lines of, "How can they possibly be going IPO? Look at their debts." Now, the debts are real. Wall Street Journal reports 600 billion commitments. The S1 filing itself shows 750 billion in commitments. Let's dissect them one by one. But before we do, here is a quick refresher to fully understand the picture. Anthropic runs on three chip families. Google TPUs, Amazon Tranium, and Nvidia's GPUs. And all of their commitments can be broken down into three groups. Commitment to Google, commitment to Amazon, and to Nvidia.对 Google 的承诺包括：向 Google Cloud 支付 1140 亿美元租赁 TPU 算力至 2033 年，加上与博通（Broadcom）的 1600 亿美元。博通是一家制造装有所有这些 TPU 芯片的物理机架的公司。总计大约 2700 亿美元，基本上占了全部承诺金额的一半。值得注意的是，Anthropic 承诺即使出现极其糟糕的情况，也会向 Google 支付这笔款项。即使 Anthropic 找不到任何可以销售的客户并且不再需要那些 TPU，他们仍必须付款。作为交换，Google 承诺建设这些产能，因为 Google 知道无论发生什么，Anthropic 都会为其买单。现在还有一个奇特的细节：除了这些，Google 同时还是 Anthropic 的投资者、竞争对手（Gemini）和供应商。接下来是对 Amazon 的承诺：向 AWS 支付 1100 亿美元以获取 Trainium 芯片和 Graviton CPU 的访问权限。最后是对 Nvidia 的承诺：向 Nscale 和 Lambda 支付 860 亿美元。Nscale 和 Lambda 是出租 Nvidia GPU 的新型云服务商（Neoclouds）。向微软支付 450 亿美元用于 Azure 算力，这些同样无论实际使用与否都必须支付。
英文原文
Commitment to Google, $114 billion to Google Cloud to rent TPU capacity through 2033 plus 160 billion with Broadcom. Broadcom is a company that creates physical racks with all those TPU chips. Roughly 270 billion in total, basically half of the entire commitment. It is interesting to note that Anthropic commits to pay this money to Google even if things go terribly wrong. Even if Atropics cannot find any customers to sell to and don't need those TPUs anymore, they still have to pay. In exchange, Google commits to building this capacity because it knows that Anthropic will pay for it no matter what. Now, a peculiar detail. On top of this, Google is also Anthropic's investor, a competitor, Gemini, and a supplier. Now, commitment to Amazon, $110 billion with AWS getting access to Tradeium chips and Gravron CPUs. And lastly, commitment to Nvidia, $86 billion with Nscale and Lambda. NScale and Lambda are Neoclouds that rent out Nvidia's GPUs. $45 billion with Microsoft for Azure compute that are again payable regardless of actual usage.算力承诺：巨额债务还是行业工具？
所有这些承诺加起来大约为 7500 亿美元。这是路透社文章中出现的数字，并在科技界被撕扯讨论。路透社因将其定性为半万亿美元的债务而受到强烈批评。我们不偏袒任何一方，但这可以有两种不同的解读：你可以将其视为一笔庞大而危险的债务，也可以将其视为打造 AI 行业巨头的一种金融工具。一方面，这些承诺中有 80% 是不可撤销的事实令人担忧。为什么？因为有证据表明，个人和企业对模型没有任何忠诚度。我可以以自己为例：我用 Anthropic 的模型处理复杂任务，用 GPT 处理视觉任务，用 Gemini Flash 处理简单问题，用结合了所有这些模型的 Perplexity 处理深度调研。我会根据任务非常轻松地在它们之间切换。这几乎就像假设你签了 10 年租约开了一家餐馆，每晚都爆满，但你所有的客人都是未经预约直接前来的散客，没有人订座。如果街对面的地方变得更便宜或更好，他们完全可能会离开。他们会离开，但你的房租依然保持不变，而且无论如何接下来的 10 年你都不得不继续支付。因此，Anthropic 已经承诺支出数百亿美元，但他们的客户却完全有可能转投另一家供应商。
英文原文
All of these commitments combined add up to approximately $750 billion. This is the number that appeared in a Reuters article and was picked to shreds across tech community. Reuters was heavily criticized for positioning this as a half a trillion dollar debt. Not to take any sides, but this can be interpreted in two different ways. You can see it as a massive dangerous debt or you can see it as a financial instrument to build a titan of the AI industry. On one hand, the fact that 80% of these commitments are non-cancellable is concerning. Why? Because there is evidence that people and businesses have no loyalty to the models. I can use my own example. I use anthropics models for complex tasks, GPT for visuals, Gemini Flash for simple questions, perplexity that combines all of them for deep research. I switch between them very easily depending on the task. It's almost like imagine you open a restaurant on a 10-year lease and it's packed every night, but all of your guests are walk-ins. Nobody has a reservation. If the place across the streets gets cheaper or better, they might as well move. They'll move, but your rent stays exactly as it was, and you're stuck paying for it for the next 10 years either way. So, Anthropic has already committed to spending billions, but it is possible that their customers are going to switch to another provider.另一方面，如果你真正阅读那篇文章，第一句话就提到 Anthropic 预计在十年内与六家合作伙伴至少支出 5180 亿美元。根据你查看的数据来源不同，10 年内 5000 亿或 7500 亿美元相当于每年 500 亿到 700 亿美元，而 Anthropic 目前的年营收为 700 亿美元。所以当你换个角度看待这些数字时，半万亿美元的债务对他们来说并没有那么夸张。Anthropic 现在进行 IPO 的原因在于，他们银行账上有大约 1300 亿美元，却有半万亿美元的账单。为了为其剩余部分提供资金，它需要现有最深厚的资本池——公开市场。那么，一家背负如此规模账单的公司打算如何偿还它们呢？要理解这一点，你需要知道如何看懂招股书。
英文原文
On the other hand, if you actually read the article, the very first sentence says Anthropic expects to spend at least $518 billion over a decade with six partners. 500 billion or 750 billion depending on which source you're looking at over 10 years is 50 to 70 billion a year and the current anthropics revenue is 70 billion a year. So when you put these numbers in perspective half a trillion debt is not that wild of a number for them. The reason Anthropic is going IPO now is because they've got about $130 billion in the bank and half a trillion dollars in bills. To fund the rest of it, it needs the deepest pull of capital there is, the public market. So, how does a company with bills that size plans to pay them? To understand that, you need to know how to read the filing.企业级 B2B 痛点与赞助商介绍
在这个频道之前，我花了 10 年时间构建科技产品。我曾在医药科技、资本市场科技和无障碍技术领域工作。这里有一个关于 B2B 的鲜为人知的尴尬秘密：数据总是一团糟。它存在于遗留基础设施上，其中一半在法律上根本无法离开大楼。客户担心模型会拿他们的数据进行训练，因此有一半时间你甚至无法发布已经构建好的功能。即使在今天，当产品交付了 AI 功能，客户虽然说“我很喜欢，我很想要，但让我先咨询一下我的法务团队”。当他们的法务团队询问数据将在哪里处理，而训练数据在美洲区域，客户的数据却在欧盟区域时，整个对话就彻底告吹了。
英文原文
Before this channel, I spent 10 years building tech products. I worked in pharma techch, capital markets, tech, and accessibility. And here is the dirty little secret about B2B. The data is always a mess. It lives on legacy infra. Half of it legally cannot leave the building. Clients are scared models are going to be trained on their data. So half the time you can't even launch what you already built. Even today when product ships an AI feature and the client says yes, I love it and I want it but let me check with my legal team first. And when their legal team asks where the data will be processed, but the training data is in the US region and the client's data is in the EU, the whole conversation is over.Cloudera 的 Anywhere Cloud 将这一问题扼杀在萌芽状态。其背后的理念是：停止将你的数据生拉硬拽到 AI 面前，而是将 AI 带到你的数据所在的地方。它可以是公共云、你自己的数据中心、主权环境甚至完全物理隔离的环境，全部遵循统一的安全与治理规则。你基本上可以在自己的安全边界内建立私有 AI 工厂，运行包括 Mistral 在内的开源权重模型，使你的数据、提示词和输出永远不会离开。它基于 Apache Iceberg 等开源技术构建，没有无休止的复制，也没有供应商锁定。没有痛苦的导入流程，因为团队可以用通俗英语提出需求并获得引导式设置，并可选择审批每一个步骤。Anywhere Cloud 让模型能够在拥有完整安全和治理控制的前提下，以可预测的成本处理你自己的数据。欢迎了解 Cloudera Anywhere Cloud，链接在简介中。现在回到视频。
英文原文
Cloudera's anywhere cloud nips this problem in the butt. The idea behind it is to stop dragging your data to the AI and bring AI to where your data already lives. It could be public cloud, your own data center, sovereign or even airgapped all under one set of security and governance rules. You can basically build your own private AI factory running openweight models including mistrols inside your own perimeter so that your data prompts and outputs never leave. It's built on open tech like Apache iceberg. There is no endless copying and no lockin. No painful onboarding because teams can ask for what they need in plain English and get guided setup with an option to approve every single step. Anywhere Cloud gets models to work on your own data with full control over security and governance at a predictable cost. Check out Cloudera Anywhere Cloud. The link is in the description. And now back to the video.解读 S-1：营收真实性与核算方式
公开的 S-1 文件自 10 月 26 日起已经公布。让我们在不深究财务分析细节的情况下阅读它。究竟什么是真正重要的？
第一点：资金是真实的吗？ 根据他们向投资者说明的数据，在今年第一季度，Anthropic 营收 47 亿美元；第二季度为 116 亿美元；第三季度为 168 亿美元。
这种增长势头意味着什么？极其迅速的增长。换个角度来看：Zoom 在疫情暴发的第一年增长了大约四倍；Nvidia 的数据中心业务增长了大约三倍；Salesforce 花了 8.5 年才实现从年营收 10 亿美元增长到 100 亿美元；而 Anthropic 在一年内将其营收增长了 12 倍。一家公司如何能实现这样的增长？当然是通过模型以及 Claude Code。Claude Code 已经成为一项独立业务，Anthropic 表示自今年年初以来它已经增长了四倍。
在所有这些数字中，值得关注的是：
- 净增额：从第一季度到第二季度增加了 70 亿美元，第二季度到第三季度增加了 50 亿美元。我所说的只是你必须关注增量金额，这在 S-1 中非常重要——是指实际增加的美元数额，而不仅仅是百分比差异。
- 资金的核算方式：Anthropic 几乎一半的营收来自 Amazon、Google 和微软的云平台。但 Anthropic 是在合作伙伴抽取分成之前，将全额售价计为自己的营收。打个比方：假设你以 100 万美元买了一套房子，但你仍欠房产中介 5% 的费用，然而你告诉全家人买下这套房子总共只花了 100 万美元，并未提及中介费用。这没有问题，完全合法，没有任何违规之处，这只是你在阅读招股书时需要注意的地方。所以，答案是：资金确实是真实的，但附带说明。
英文原文
The public S1 has been out since October 26th. Let's read it without getting into the depths of financial analysis. What is actually important? Number one, is the money real? In the first quarter of this year, Anthropic made $4.7 billion. In the second, 11.6. In the third, 16.8. According to what they said to their investors.What does this progression mean? very rapid growth. To put this in perspective, Zoom in the first year of COVID grew about four times. Nvidia's data center business grew about three times. Salesforce took 8.5 years to go from $1 billion to 10 billion a year. Anthropic grew its revenue 12 times in one year. How does a company grow like that? Models of course and cloud code. Cloud code became its own business and a tropic said that it had quadrupled since the start of the year. Now of all of these numbers what is worth watching one the gain Q1 to Q2 added 7 billion Q2 to Q3 five all I'm saying is that you have to watch the gain it is important in the S1 the gain as in the dollars added not just the percentage difference two how the money is counted almost half of entropics revenue comes through Amazon's Google's and Microsoft's cloud But Anthropic counts the full price as its own revenue before the partners take the cut. The analogy here, imagine you buy a house for a million dollars, but you still owe your real estate agent 5%. But you told your whole family that all it cost you to get the house is a million, not mentioning the fees. This is fine. It's totally legal. There is nothing wrong about it. It's just something you have to watch when you read the filing. So yes, the money is real with an asterisk.
解构 807 亿美元净亏损的真相
第二点：亏损是真实的吗？ 另一个被广泛讨论的数字是：2026 年上半年净亏损 807 亿美元。现在，我想请大家注意一点。如果你打开报道这一数字的新闻文章本身，第一句话就解释了这笔亏损的来源：“在很大程度上归因于可转换为该 AI 初创公司股票的债券公允价值上升”。
这意味着什么？早期，一些投资者借钱给 Anthropic。他们不是以现金偿还，而是将该笔贷款以固定价格换取 Anthropic 的股票。这种类型的贷款被称为可转换债券。会计准则规定，一家公司不能永远按原先的初始金额记录该笔贷款。为什么？因为每个季度如果公司增长，股票也会随之增值，价值更高；因此，他们所承担的贷款价值也更高。这些规则的存在是为了确保当公司股票升值时，投资者借给该公司的资金也能相应增值。
在 2026 年上半年，Anthropic 的估值飙升。因为估值飙升，贷款的价值也水涨船高，猛增了 720 亿美元。这意味着由于公司价值上升，Anthropic 欠其投资者的债务增加了 720 亿美元。而这 720 亿美元的账面变动正是所报道的 807 亿美元亏损的大部分来源。他们并没有亏损这笔资金，他们只是因为履约责任而必须向投资者支付。
文章本身就提供了证据：“其上半年亏损从一年前的 130 亿美元进一步扩大，其中包括一笔 720 亿美元的会计费用，反映了最终可转换为 Anthropic 股份的债券预估价值的增长。”这种亏损仅仅表明那些早期贷款人最终将拥有公司大得多的份额。这就是所谓的亏损。
英文原文
Two, is the loss real? Another widely discussed number, $80.7 billion net loss in the first half of 2026. Now, I want to bring your attention to something. If you open the news article itself that reported this number, the very first sentence explains where this loss is coming from. in large part due to a rise in the fair value of bonds convertible into shares of the AI startup. What does this mean? Early on, some investors lent anthropic money. Instead of being paid back in cash, they swapped that loan for shares of Anthropic at a fixed price. This type of loan is called a convertible bond. Accounting rules say that a company cannot have that loan at that original amount forever. Why? Because every quarter if the company grows, so do the shares. They're worth more. And therefore, the loan that they took is also worth more. These rules exist so that when the shares of the company go up in value, the investors money that they had lent to the company appreciate too. In the first half of 2026, Anthropics value shot up. Because it shot up, the value of the loans also went up and they jumped $72 billion. Meaning that Anthropic owes its investors $72 because the value of the company went up. And $72 billion loss is much of that 80.7 reported. They didn't lose that money. They simply had to pay their investors because it's their obligation. Here's the proof in the article itself. Its first half loss, widening from a loss of 13 billion a year earlier, included a 72 billion accounting charge, reflecting an increase in the estimated value of bonds that could eventually convert into anthropic shares. This loss only shows that those early lenders will end up owning a much bigger slice of the company. That's the loss.合同营收与客户忠诚度风险
这就引出了第三点：他们真的有可能亏钱吗？ 依我浅见，有一个数字比 80 亿美元（注：此处指巨额亏损）更值得关注得多——Anthropic 的合同约定营收（Contracted Revenue）为 108 亿美元。合同约定营收是指一家公司理应获得但尚未实际收到的收入。而且这笔合同约定营收中的 75% 是在过去 12 个月内签订的。现在把这个数字与 Anthropic 在 2026 年第二季度实现的 116 亿美元相比。他们作为一个企业整体所签约的营收，甚至无法覆盖两个季度的销售额。
这几乎就像你决定辞去全职工作，因为你的副业正在蓬勃发展，但你拥有的现金周转期（Runway）还不到两个季度，甚至这两个季度的周转期还是你指望之前雇佣你的公司支付给你的款项，你手里还没有拿到钱，但你明天就要辞职了。
我为什么觉得这令人担忧？因为 Anthropic 80% 的营收来自按量付费。根据 S-1 文件，这部分营收中的大部分并未绑定长期合同，与他们签订合同的客户可以随时削减或停止支出。想一想这意味着什么：这意味着合同约定营收以及他们预期的未来营收非常不稳定。客户对 Anthropic 没有忠诚度，他们只忠诚于谁能以最合理的价格提供最好的模型。我们已经可以看到，作为 Anthropic 最大的客户之一，Cursor 在 8 月份被 SpaceX 收购。本月，Cursor 推出了捆绑 Grok 的新套餐，而 Claude 则需要额外收费。
对此你可能会说，花在 AI 上的钱只会越来越多。你说得对。去年留在 Anthropic 的客户花费的金额大约增加了九倍。但是看看代币（Token）价格发生了什么变化：Anthropic 最新的 Opus 执行相同任务的费用比上一代降低了大约 40%。S-1 本身也发出警告：竞争对手可能会以更低的成本提供足够好的性能。那么，他们可能会亏钱吗？是的，这完全有可能，因为让他们保持在行业最前沿的护城河，是拥有最优秀的编程模型。而代码编写是 AI 最大的用例。但编写代码并不是一切。总而言之：钱是真的吗？是的，钱是真的。他们会亏钱吗？是的，他们绝对有可能亏钱。
英文原文
Which leads me to number three. Can they actually lose the money? There is a number that in my humble opinion deserves far more attention than the $8 billion loss. Anthropics contracted revenue 10.8 billion. Contracted revenue is the revenue that a company is supposed to earn but hasn't earned yet. And 75% of this contracted revenue was contracted within the last 12 months. Now compare that number to the 11.6 billion that Anthropic made in Q2 2026. The revenue that they have contracted as a whole, as a business, doesn't even cover two quarters of sales. It's almost like if you decide to leave your full-time job because your side business is taking off, but you've got less than two quarters of runway and even those two quarters of runway is what you expect the company that you worked for to pay you, but you don't have the money yet, but you're quitting tomorrow. Why do I find it concerning? Because 80% of Anthropics revenue is pay-per- use. And according to S1, much of that revenue isn't tied to long-term contracts, and the customers that they have contracts with can reduce or stop spending at any time. Think about what this means. It means that the contracted revenue and the future revenue that they expect is very wobbly. Customers are not loyal to Anthropic. They're loyal to whoever has the best model for the most reasonable price. And we can already see that cursor, one of Anthropic's biggest customers, was bought by SpaceX in August. This month, Curser launched new plans that bundle Grock and Claude comes at an extra cost. Now, to that you may say the amount of money spent on AI is only growing. And you'd be right. Customers who were with Anthropic last year spend about nine times as much. But look at what's happening with token prices. Antropic's newest Opus does the same task for about 40% less than the one before it. And the S1 itself warns that competitors may offer adequate performance at a lower cost. So, can they lose money? Yes, it's very possible because the thing holding them at the forefront of the industry is having the best model for coding. And coding is the number one use case for AI. But coding isn't everything. So to sum it up, is the money real? Yes, the money is real. Can they lose the money? Yes, they absolutely can.上市对日常用户与投资者的影响
这就引出了最后一点：当 Anthropic 成为一家上市公司后，你的生活会发生什么改变？
如果你在科技行业工作，我不知道你们公司内部是否举办过关于“代币经济学”的网络研讨会。很多公司现在都在办这种培训，基本上是在教育员工如何优化使用 AI，比如像“去转角的咖啡馆不需要开法拉利”，如果你只有一个非常简单的问题，就不要选择最先进的模型。但如果你站在如今必须向股东交代的 Anthropic 的立场上，你的利益诉求恰恰相反：你希望从每位客户身上尽可能多地获取收入。
因此，一个合乎逻辑的问题是：价格会上涨吗？ 不一定，但要关注使用限额（Limits）。Anthropic 可能会调整限额，而价格保持不变。当我提到“限额”这个词时，我指的是非常宽泛的范畴：它可以是套餐中包含的工作量配额；可以是模型消耗的额度；也可以是额外代币或额外使用量需要开始单独计费的阈值。而且这已经有先例了：首先，在 5 月至 9 月期间，Claude Code 将每周限额提高了 50%；随后在 9 月，它将其中 25% 的增幅固定下来。现在，如果限额发生变化，对企业的影响很可能会远大于个人订阅者。如果你是企业主，或者如果你所在的公司使用 Claude 处理工作，你们公司并不是为每个工位支付固定费用，而是按你使用 Claude 的实际用量付费。因此，限额的任何变动都会直接反映在你们公司的账单上。
如果你进行投资：如果你的 401(k)、RRSP 或 ISA 养老金账户中持有纳斯达克 100 指数基金，你可能会在几周内直接间接持有 Anthropic 的股票。今年早些时候，纳斯达克修改了规则，因此一家新上市的巨头在交易仅两周后即可被纳入纳斯达克 100 指数。标普 500 指数则有所不同：它要求公司上市满 12 个月并实现真实盈利。对于标普 500 而言，800 亿美元的账面亏损正是将 Anthropic 挡在门外的数据。同样，情况可能会改变，但所需时间将远超两周。
所以，这是我的个人结论：营收是真实的。去年几乎增长了 12 倍，这个数字白纸黑字写在 S-1 文件中。所报道的亏损主要是由会计规则导致的。然而，尚未明确的是高度集中的收入结构和缺乏客户忠诚度的现状，这意味着 Anthropic 确实在进行一场巨大的豪赌——押注在未来十年中始终拥有最好的模型。
接下来需要关注的三件事：
- 第一份财报；
- 锁定收入的 108 亿美元是否开始增长；
- 使用限额的变动。
让我们看看是否会有所改变。请在评论区告诉我们你对这次历史性 IPO 有什么期待，我们每一条都会认真阅读。感谢收看，我们下期再见。
英文原文
Which leads me to the last point. What will change in your life when Anthropic becomes a public company?If you work in tech, I don't know if you've had one of those tokconomics webinars at work. A lot of companies are doing them right now where they're basically educating people on how to use AI optimally. things like you don't need a Ferrari to get to the closest coffee shop around the corner. So don't pick the most advanced model if you have a very simple question. But if you put yourself in the shoes of Anthropic that now has to answer to shareholders. Your interests are exactly the opposite. You want to earn as much as possible per customer. Therefore a logical question would be will the prices go up? Not necessarily but watch the limits. Anthropic can change the limits, but the price may stay the same. When I say the word limits, I use it very loosely. It could be how much work is included in the package. It could be how much allowance models consume. It could be the point when additional tokens or additional usage becomes separately billable. And there is a precedent already. First, between May and September, clot code increased weekly limits by 50%. Then in September, it made a 25% increase permanent. Now, if the limits change, this will most likely affect businesses much more than individual subscribers. If you're a business owner or if you work for a company that uses Claude for work, your company doesn't pay a flat fee for your seat. It pays for how much you use Claude. So, any change in the limits will show up on your company's bill if you invest. If you have a NASDAQ 100 fund in your 401ks, RRSP or ISA, you may own Anthropic directly within a couple of weeks. Earlier this year, NASDAQ changed the rules. So, a giant new listing can join the NASDAQ 100 after just 2 weeks of trading. S&P 500 is different. It requires 12 months as a public company and a real profit. And for S&P 500, the $80 billion loss is the kind of number that keeps Anthropic out of the S&P 500. Again, it may change, but it'll take way more than 2 weeks. So, here is my personal verdict. The revenue is real. It grew almost 12fold last year, and that number is in S1. The loss that's been reported is mostly accounting rules. What isn't settled though is the concentration of revenue and absence of loyalty which means that Anthropic is indeed making a massive bet on having the best model for the next decade. Three things to keep an eye on next. The first earnings report, whether that 10.8 billion of locked in revenue starts growing and the limits. Let's see if something changes. Tell us in the comments what you're going to be expecting from this historic IPO. We will be reading every single one. Thank you for watching. Till next time.